Extensive Editing of mRNAs for the Squid Delayed Rectifier K+ Channel Regulates Subunit Tetramerization

Neuron, Vol. 34, 743–757, May 30, 2002, Copyright 2002 by Cell Press

Extensive Editing of mRNAs for the Squid Delayed Rectifier Kⴙ Channel Regulates Subunit Tetramerization Joshua J.C. Rosenthal and Francisco Bezanilla1 Departments of Physiology and Anesthesiology UCLA School of Medicine Los Angeles, California

Summary We report the extensive editing of mRNAs that encode the classical delayed rectifier Kⴙ channel (SqKv1.1A) in the squid giant axon. Using a quantitative RNA editing assay, 14 adenosine to guanine transitions were identified, and editing efficiency varied tremendously between positions. Interestingly, half of the sites are targeted to the T1 domain, important for subunit assembly. Other sites occur in the channel’s transmembrane spans. The effects of editing on Kⴙ channel function are elaborate. Edited codons affect channel gating, and several T1 sites regulate functional expression as well. In particular, the edit R87G, a phylogenetically conserved position, reduces expression close to 50-fold by regulating the channel’s ability to form tetramers. These data suggest that RNA editing plays a dynamic role in regulating action potential repolarization in the giant axon. Introduction RNA editing, mediated by the enzymatic conversion of adenosine to inosine (A→I), regulates nervous system function in important ways. First identified in Xenopus laevis (Bass and Weintraub, 1987, 1988), and subsequently verified in a diverse variety of organisms, A→I conversion appears to be ubiquitous among metazoans (Polson et al., 1991). Based on the abundance of inosine in mRNAs, it has been suggested that a surprisingly large percentage of brain mRNAs may be edited (Paul and Bass, 1998). In spite of this, few specific substrates for A→I conversion have thus far been identified, and the functional consequences of editing are understood in only a handful of these cases. At present, the biological significance of editing is uncertain. Is the primary purpose of RNA editing to expand an organism’s genomic repertoire? Are editing sites selected on the basis of an individual protein’s function, or do they affect diverse proteins in common ways? Answers to these questions require diverse examples of how editing regulates protein function. In vertebrates, substrates for RNA editing have been identified from the mammalian brain and encode proteins involved in synaptic transmission. The most thoroughly studied examples are the mRNAs encoding GluR-B and other glutamate receptor channel subunits, where editing regulates the receptor’s calcium permeability and gating kinetics (Sommer et al., 1991; Kohler et al., 1993; Lomeli et al., 1994). More recently, it was reported that editing of a serotonin receptor subunit’s 1

Correspondence: [email protected]

mRNAs (5HT2C) reduces the receptor’s affinity for its G protein (Burns et al., 1997). Editing of several viral RNAs which infect mammals is also well characterized (Bass et al., 1989; Polson et al., 1996). A→I editing is carried out by a group of enzymes known collectively as ADARs (Kim et al., 1994; Melcher et al., 1996a, 1996b; O’Connell et al., 1995). ADARs bind to regions of double-stranded pre-mRNA and catalyze the hydrolytic deamination of adenosine. Higher organisms have multiple ADARs, and the process of editing is complicated by the fact that a single mRNA can be edited at different sites by different enzymes (Melcher et al., 1996b; O’Connell et al., 1997). Comparatively little is known about A→I editing in invertebrates, but here too the nervous system is targeted. In several species of Drosophila, transcripts from the Para locus, which encodes the major brain Na channel, are edited (Hanrahan et al., 1999, 2000). Recent reports indicate that a Drosophila melanogaster Ca2⫹ channel (Smith et al., 1998) and a glutamate-gated chloride channel (Semenov and Pak, 1999) are also edited. Functional data are not available for these examples. An interesting feature of editing in invertebrates is that, in some cases, a high percentage of adenosines within a single mRNA can be deaminated, a phenomenon known as hyperediting. For example, in transcripts from the Drosophila melanogaster 4F-rnp locus, up to 31% of adenosines are thought to be converted to inosine (Petschek et al., 1996, 1997). In a Kv2 K⫹ channel (SqKv2), expressed in the brain of the squid Loligo pealei, 17 editing sites occur within a 360 nt segment (18% of the available adenosines; Patton et al., 1997). Two of these sites influence the channel’s rate of closing and inactivation. Thus it appears that in Loligo, as well as in the mammalian brain, A→I editing affects the machinery for impulse propagation. Further, the high incidences of editing in the examples noted above suggest that A→I conversion is a particularly robust process in invertebrates. The selection of K⫹ channel mRNAs as a target for editing in squid brain is particularly interesting in light of the many critical functions that these proteins regulate (Pongs, 1999). A neuron’s resting potential, firing threshold, firing frequency, as well as the rate of action potential repolarization, are all regulated by K⫹ channels. Voltage-dependent K⫹ channel ␣ subunits of the Kv1-4 families form tetrameric proteins (MacKinnon, 1991) which open a K⫹-selective pore in response to depolarization. In these channels, the mechanisms for voltage sensing (Bezanilla, 2000) and ion permeation (MoraisCabral et al., 2001) have been rigorously studied and are well understood. An amino-terminal tetramerization domain (T1) is also well characterized. By binding to T1’s in other monomers, this domain is instrumental for subunit tetramerization during channel maturation (Li et al., 1992, 1993), and regulates the formation of heteromultimers. Because they contain compatible T1 domains, ␣ subunits within the same subfamily (e.g., Kv1) can freely coassemble, whereas those between subfamilies cannot (Li et al., 1992; Shen et al., 1993; Xu et al., 1995). Crystal structure data have revealed that four T1

Neuron 744

domains form a symmetrical structure surrounding a water-filled cavity, and that the amino acids at the interface between subunits are highly polar (Bixby et al., 1999; Kreusch et al., 1998; Minor et al., 2000). K⫹ conductance (gK) in the squid giant axon has been intensively studied for almost 50 years and has led to much of what we know regarding how K⫹ channels operate. On a molecular level, a Kv1 ␣ subunit (SqKv1.1A) has been implicated as the major channel responsible for action potential repolarization in this system (Rosenthal et al., 1996, 1997). Interestingly, significant levels of A or G variation between individual SqKv1.1A cDNAs were identified (Rosenthal et al., 1997). We decided to examine whether SqKv1.1A mRNAs were edited as well as those for SqKv2 because the giant axon system offers numerous practical advantages for studying how molecular-level changes can regulate physiological properties. In this one preparation, single channel currents, gating currents, and macroscopic ionic current flowing during an action potential can all be measured (Armstrong and Bezanilla, 1973; Bezanilla et al., 1970a, 1970b; Llano et al., 1988; Vandenberg and Bezanilla, 1991). In their landmark study, Hodgkin and Huxley demonstrated that a single particle (N), with an open probability raised to the fourth power, could adequately describe gK in the giant axon (Hodgkin and Huxley, 1952). In the present work, we report that the molecular basis of gK in this system is substantially more complex. Extensive editing of the SqKv1.1A mRNA yields a variety of ␣ subunits with altered biophysical properties. A novel finding is that numerous editing sites are clustered in sequence encoding the T1 domain, an area not previously thought to be a target for channel regulation. Selective editing of these sites modifies the channel’s ability to oligomerize, and, as a consequence, regulates total potassium conductance. Taken together, these editing events provide a potent mechanism for regulating total potassium conductance over a considerable voltage range. Results Genomic Structure and Copy Number of SqKv1.1 In a previous study (Rosenthal et al., 1997), four related squid Kv1 family cDNAs were compared on a molecular level (SqKv1.1A-D). These clones were highly similar throughout their core and 3⬘ ends, but diverged completely at their 5⬘ ends. A close inspection of the core, however, revealed 10 positions with A or G variability. Sequence analysis of multiple cDNA clones from SqKv1.1A and B revealed further A or G variability within each class. Based on these findings, it was hypothesized that the divergent 5⬘ ends were due to alternative splicing, and the A or G variability was due to RNA editing. However, alternative explanations were plausible. Both differences could be explained by multiple, closely related genes. In addition, because the cDNA library used in the study was constructed by pooling tissue from several hundred squid, population-level variation was an additional concern. Therefore, to determine whether A or G variability was indeed due to RNA editing, these questions had to be addressed. To answer both questions, it was necessary to isolate the genomic se-

quence for the SqKv1.1 locus and to assess A or G variation within individual squid. Alignments of the putative splice variants SqKv1.1A-D (Rosenthal et al., 1997) suggested that an intron was likely to occur in the SqKv1.1 locus near the 5⬘ end of the coding sequence (between nucleotides 35 and 36 of the SqKv1.1A cDNA; upside down triangles in Figure 1). This prediction was supported by several PCR trials using squid genomic DNA as template. Amplifications using primers that spanned the putative intron site were unsuccessful. However, amplifications using primer pairs on either side of this site were successful and yielded products of the size predicted from the cDNA sequence. In fact, after nt 36, the entire coding sequence could be isolated from a single amplification and contained no introns. The exact position of the SqKv1.1 intron was verified by cloning its 5⬘ and 3⬘ ends (Figure 1, also see Experimental Procedures). As expected, genomic sequence revealed perfect splice donor and acceptor motifs for eukaryotic introns between the equivalent of nt 35-36 of the cDNA sequence (Figure 1B, shaded residues; Padgett et al., 1986). Thus, the genomic structure of the SqKv1.1 locus’ coding region which underlies the SqKv1.1A mRNA contains 2 exons bounding a single intron. The 5⬘ exons, which underlie SqKv1.1B-D, have not yet been identified. Southern blots of squid genomic DNA were performed to determine the SqKv1.1 copy number (Figure 1C). A 32 P-labeled SqKv1.1 DNA probe was synthesized from an exon 2 restriction fragment which included most of the channel’s core. Blots using squid genomic DNA cut by restriction enzymes with sites at known positions in the SqKv1.1A cDNA, or within the intron identified above, yielded single bands of the predicted size (Figure 1Ci). This verified the finding that this gene contains a single intron. Further, blots were carried out using enzymes that do not cut within SqKv1.1A (Figure 1Cii). Because untranslated regions flanking a gene are generally highly variable, restriction enzyme sites would not be expected to be preserved between different gene copies. Therefore, the presence of multiple SqKv1.1 genes should be revealed by multiple bands on the blot in Figure 1Cii. However, for each of the four enzymes tested, a single band was evident. These results suggest that a single gene underlies SqKv1.1A. The genomically encoded sequence for the entire coding region of the SqKv1.1 locus (Figure 3), SqKv1.1G, was derived from the experiments described above and verified repeatedly by the editing assay described in the following section. Although highly similar to SqKv1.1A, SqKv1.1G differed at 9 positions (nt 107, 127, 259, 418, 523, 535, 766, 780, and 1129). At each position, SqKv1.1G had an A and SqKv1.1A had a G. These sequences, however, were derived from different individuals and therefore population-level genetic variation was still an alternative explanation to RNA editing. In addition, if editing was in fact occurring, the SqKv1.1A cDNA first identified was not necessarily edited at all possible sites. To specifically address these questions, we developed a direct sequencing RNA editing assay. A Direct Sequencing RNA Editing Assay Our assay was developed for two purposes: (1) to identify editing sites and (2) to estimate the editing frequency

RNA Editing Regulates K⫹ Channel Tetramerization 745

Figure 1. Genomic Structure of the SqKv1.1 Locus (A) A schematic of the SqKv1.1A cDNA, showing approximate positions of features discussed in text. Open box represents coding region. Filled line represents untranslated regions. Arrows show positions of restriction sites. Upside down triangle indicates position of the single intron discovered for this gene. (B) Blow up of region surrounding intron. Top line is from cDNA sequence and upside down triangle indicates intron’s position. Bottom two lines are same region derived from genomic DNA. Vertical dashed lines indicate precise boundary with intron. Gray highlighted residues indicate conserved splice acceptor and donor sites. (C) Southern blots using L. opalescense genomic DNA cut with the indicated restriction enzymes. Molecular weights were taken from markers run on the gels.(i) Double cuts. In the SqKv1.1A cDNA, restriction enzymes cut at the following positions: NdeI- nt 101, AccI- nt 1018, SpeInt 1145, BstyI- in intron 1 644 nt from 3⬘ end. Sizes on the blot are predicted by these positions. (ii) Single cuts with enzyme that do not cut in the SqKv1.1A cDNA. Only single bands were evident. For all blots, a 32P-labeled randomly primed DNA probe was synthesized from the NdeI-AccI restriction fragment.

at these sites. Moreover, because only a tiny amount of giant fiber lobe (giant axon somata) RNA can be isolated from a single squid, the assay had to be sensitive. To meet all of these requirements, cycle sequencing, using 33 P-labeled ddNTPs (Thermo Sequenase Radiolabeled Terminator Cycle Sequencing Kit, Amersham) was employed. This chemistry not only was sensitive enough to identify low abundance editing events, but it also could be used to estimate frequencies because it produces even and consistent band intensities on a sequencing gel. Because this approach to identifying editing sites has not been previously reported, its utility is demonstrated in detail in Figure 2. Figure 2A shows eight consecutive bands in a 33P-ddGTP terminator reaction (G lane) from the sequence of a SqKv1.1G template. A casual inspection of the bands indicates that their intensities are fairly even. This sequencing reaction was repeated five times and the resulting bands were analyzed on a phosphorimager. In a single reaction, the standard deviation (SD) of the intensities of all eight

bands was ⫾11% on average (n ⫽ 5). However, the relative intensity of each specific band was remarkably constant between reactions. Each bar in Figure 2B marks the intensity of an indicated band relative to the average intensity of all eight bands for one reaction: clearly the relative intensity of each band is very consistent between different reactions. The relative intensities for band 1, the least consistent of the eight, varies by ⫾4.7% (SEM, n ⫽ 5) whereas the relative intensities of band 5, the most consistent, varies by ⫾2.2% (SEM, n ⫽ 5). On average, the variability of relative intensities for the eight bands shown is ⫾2.9% (SEM); this systematic error is a measure of the precision of the assay and was considered in all subsequent analysis. In Figures 2C and 2D, the accuracy of our assay for determining the frequency of editing at a site is tested by characterizing the quantitative relationship between a band’s intensity and template’s abundance. Figure 2C shows sequence which has been amplified from two closely related clones. These clones differ at position

Neuron 746

Figure 2. Direct Sequencing Editing Assay (A) Phosphorimager scan of a single lane of sequence for 33P-labeled terminator chemistry (A reaction, SqKv1.1G plasmid template). (B) Reaction in (A) repeated five times. For each reaction, to calculate the fractional intensity, the intensity of an individual band was divided by the average intensity of all bands. These ratios were plotted for each band. Each bar represents one of the five trials. See Experimental Procedures for details on quantification. (C) Controls for quantification of mixed position. SqKv1.1G or SqKv1.1A were sequenced and the vicinity of nt 535 is shown. This nt (arrow 1) is an A in SqKv1.1G and a G in SqKv1.1A. Only the A and G reactions are shown. The first two templates are SqKv1.1G and SqKv1.1A, respectively. For the subsequent reactions, the two plasmids have been mixed at the indicated molar ratios. (D) Analysis of part (C). For A lanes, the ratio of band 1 to band 2 was calculated (filled circles) and for G lanes, the ratio of band 1 to band 3 was calculated (open triangles). These ratios were plotted against the plasmid template ratio. (E) Schematic of editing assay extended to Loligo. Amplification products used as templates for direct sequencing were 1466 nt for genomic DNA reactions and 1460 nt for cDNA reactions. See Experimental Procedures for specific primers. (F) Examples of two edited positions in SqKv1.1 (nt 63 and 780). Arrows indicate edited positions.

1, where the “genomic” clone has an A and the “cDNA” clone has a G. The ratio of band 1 to band 2, or of band 1 to band 3, was then calculated for the genomic A reaction and the cDNA G reaction. In the right panel of Figure 2C, the genomic and cDNA clones were mixed prior to amplification and sequencing, and the molar ratios of the mixes (genomic:cDNA) are listed at the top of the panel. Clearly, the intensity of position 1 bands in the A lanes is related to the abundance of the genomic clone, and in the G lanes to the abundance of the cDNA clone. This observation is quantified in Figure 2D. Here the intensity of band 1 relative to band 2 or 3 (after being normalized to the relative intensities of unmixed clones determined in Figure 2C, left panel) is plotted versus the ratio of genomic:cDNA clone in the mix. In all cases, the experimentally determined data give a good estimate of the plasmid mix. Estimates are more accurate for relatively intense bands. Faint bands tend to produce a slight overestimate. Our direct sequencing approach was then extended to SqKv1.1 from individual squid (Figure 2E). Both mRNA, extracted from the GFL neurons of the giant axon system, and genomic DNA were isolated from the same animal. After synthesizing cDNA from the mRNA, both the cDNA and genomic DNA were used as templates for

PCR amplification of the entire coding region of SqKv1.1. Oligonucleotide primers were then used to sequence both PCR products as described above. At 14 sites where the genomic DNA sequence contained only an A, the cDNA sequence gave evidence of a G. There were no other discrepancies between the sequences. Figure 2F shows two examples of these positions. At nucleotide 63, the genomic reaction clearly has a band in the A lane and nothing in the G lane. At the same position in the cDNA reaction, however, there are bands in both lanes. At position 780, the pattern is somewhat different. In this case, as before, there is an A in the genomic DNA reaction. In the cDNA reaction, however, there is little trace of an A and a dominant band in the G lane. Both positions were quantified as described above. A G was present in the cDNA lane 65% of the time at nt 63, and 97% of the time at nt 780. This analysis was repeated at all 14 sites. Editing Site Map for SqKv1.1A mRNAs Figure 3 presents the genomically derived coding sequence for SqKv1.1G, the positions of all 14 sites of A or G variation between this sequence and cDNA sequences, and the conversion frequency at each site. Because of the small systematic errors associated with

RNA Editing Regulates K⫹ Channel Tetramerization 747

Figure 3. Editing Map of SqKv1.1G Genomic sequence of SqKv1.1G (intron excluded) with predicted protein sequence written below the first position of each codon. Edited positions are boxed. If the position causes an amino acid change, then it is indicated below. The number above each editing site refers to the relative frequency of editing (1 ⫽ 1%–25%, 2 ⫽ 26%–50%, 3 ⫽ 51%–75%, 4 ⫽ 76%–100%). Regions of known functional significance are indicated in gray. T1 ⫽ tetramerization domain. S1–S6 ⫽ membrane spans 1–6. P ⫽ pore. Data are derived from four individual squid.

our approach, and the small differences between individual squid, these results were binned in increments of 25% (i.e., 1 ⫽ 1%–25%, 2 ⫽ 26%–50%, 3 ⫽ 51%–75%, 4 ⫽ 76%–100%). Clearly, the extent of editing varies greatly between sites. Although 9 of 14 sites are converted at the highest frequency, examples of each category can be found. All sites from the channel core are edited at high frequency. Interestingly, only one site (in codon E21-nt63) is silent. Of the remaining sites, many create conservative amino acid changes (between I, V, and M). Some, however, create less conservative changes. Examples of moderate (N45S) and large (Y36C) changes in R group size are evident. In codon 87 (R87G), a charge is neutralized. In codon 132, both the first and second positions can be edited, yielding three possible changes: a charge reversal (K132E), a charge neutralization (K132G), and a charge maintenance (K132R). In general, the edited positions are clustered in sequence coding for two regions of SqKv1.1A: the tetramerization domain (T1) (Li et al., 1992) and, to a lesser extent, the transmembrane spans. Seven sites lie within the T1 domain (Figure 3). The genomically encoded amino acid for R87 is very highly conserved between

all members of the 4 Kv1-4 subfamilies, and lies at the end of an extremely highly conserved motif (NEYFFDR). Position N45 is absolutely conserved. The equivalents of both positions in an Aplysia Kv1 channel, for which the T1’s crystal structure has been solved, lie at the intersubunit interface (Bixby et al., 1999; Kreusch et al., 1998). Outside the T1 domain, four edited sites occur within membrane spans (Figure 3), two in S1 (M175V and I179V), and two in S3 (I256V and I260M). The equivalent of all four positions in the Shaker K channel are predicted to face the lipid bilayer (Li-Smerin et al., 2000; Hong and Miller, 2000). No sites occur within the highly conserved motifs of the gating machinery (S4) or the pore. Although close to the pore, I377V is a highly variable position among Kv channels and is predicted to be in the outer vestibule (Doyle et al., 1998). Edits in the Channel’s Core Affect Voltage Sensitivity and Closing Kinetics The remainder of this study focuses on the functional consequences of editing the SqKv1.1 sequence at the sites identified above. These experiments were carried out using voltage-clamped Xenopus oocytes, and, in

Neuron 748

the case of certain T1 edits, using biochemical assays. As a first step, fully edited (SqKv1.1C; G at all 14 sites), and fully genomic (SqKv1.1G: A at all 14 sites), versions of the channel were constructed. The construct SqKv1.1C, however, failed to express detectable currents in oocytes. Subsequently, sites in the channel’s core (S1-S6) were examined separately from those in the T1 domain. In general, single mutations introduced into the SqKv1.1G background were used to assess the functional effects of each editing site. Currents from channels encoded by SqKv1.1G cRNA have general properties similar to those described for SqKv1.1A (Liu et al., 2001; Rosenthal et al., 1996, 1997). Representative K⫹ current traces, in response to 25 ms voltage steps to a variety of potentials, are illustrated in Figure 4A. SqKv1.1G channels activate at voltages more positive than ⵑ⫺30 mV. Like most Kv1 family members, they open quickly in response to strong depolarizations, and voltage sensitivity is steepest between ⫺30 mV and ⫹10 mV. Little to no inactivation occurs on this time base. Over longer time scales (⬎ⵑ100 ms), slow inactivation is apparent, but this property was not examined in this work. Channel closing, like opening (Figure 4A, in response to repolarization to ⫺90 mV), is also fast. Mutations studied in this paper affected several of these properties. To examine the combined effects of all five core editing sites, K⫹ currents (Ik) from SqKv1.1G and SqKv1.1CE (M175V, I179V, I256V, I260M, I387V) were compared. IK from these constructs differ in two ways (Figure 4B). First, although both channels begin to open at a similar potential, their steady-state conductance versus voltage relationships (g-V) have different shapes. The g-V for SqKv1.1G (filled circles) is relatively steep, having activated to 90% of the peak value by ⫹10 mV. For SqKv1.1CE (open circles), 90% activation is not reached until ⵑ⫹40 mV. The difference between the two g-V curves cannot be explained by a simple shift along the voltage axis; it is their shapes that differ, a result consistent with a difference in late steps along the activation pathway. The channel constructs also have different closing kinetics. At all potentials, SqKv1.1CE closes faster, and the relationship between its deactivation time constant and voltage (␶off -V) is shifted ⵑ30 mV toward depolarized potentials. Despite this shift, the shape of the ␶off -V curves are similar for these two channels. Are the functional differences between the unedited channel and the fully core edited channel caused by a single edit, or the sum of multiple edits? To answer this question, single mutations were introduced into the SqKv1.1G background. Only those mutations that result in significant changes in channel function are reported. The SqKv1.1G I260M edit has the greatest single effect, and can account for a large portion of the changes in both the g-V curve and ␶off -V curve noted above (Figure 4C). The double mutant SqKv1.1G M175V I179V causes a similar, albeit smaller, effect on both properties (Figure 4D). M175V or I179V, when introduced individually, cause no significant effect, a result not surprising considering their comparatively small combined effect. The I387V edit, while having no effect on the g-V relationship, slowed the ␶off -V relationship significantly (Figure 4E).

Figure 4. Electrophysiological Characterization of Core Edits (A) Currents recorded from Xenopus oocytes expressing SqKv1.1G clamped with the COVG method. Oocytes were held at ⫺80 mV and then stepped to the indicated voltages for 25 ms. See Experimental Procedures for solutions. (B) gK versus voltage relationship and ␶off versus voltage for core edit mutants. gK was determined by normalizing tail current amplitude at each voltage to maximum tail current amplitude for all voltages. ␶off was measured by fitting a single exponential (see Experimental Procedures). Oocytes were pulsed to various negative values following an activating pulse to ⫹50 mV for 10 ms. The first graph for both gK versus V and ␶off versus V include the constructs SqKv1.1G (filled circles) and SqKv1.1CE (M175V, I179V, I256V, I260M, I387V, open circles). For subsequent graphs, data from these two constructs are replotted in addition to that from the indicated mutant. All experiments performed at 20⬚C. Analog signals were filtered at 10 kHz. Error bars (SEM) are included when they are bigger than the symbols.

RNA Editing Regulates K⫹ Channel Tetramerization 749

The Effects of T1 RNA Editing Sites on Voltage Sensitivity and Closing Kinetics Editing sites in the channel’s T1 domain were studied in the same manner as described above. First, all T1 edits (M35V, Y36C, I43V, N45S, R87G, K132E) were introduced into the SqKv1.1G background simultaneously. This construct failed to express detectable currents. Edits were then introduced individually. Of these, SqKv1.1G N45S (filled squares) and SqKv1.1G K132E (open triangles) produced channels whose functional properties were significantly different than those of SqKv1.1G (Figure 5A). Both constructs begin to activate at more depolarized potentials than the unedited channel, and the entire g-V relationship is shifted rightward along the voltage axis. Depending on the voltage, the magnitude of this shift varied from 10–20 mV. Channel closing kinetics were also made faster by both T1 mutations (Figure 5B). For other T1 edits (except SqKv1.1G R87G), these properties were not statistically different. Functional properties of SqKv1.1G R87G were not examined in detail because the expression level of this construct was too low to permit high quality measurements. The Effects of RNA Editing Sites in the T1 Domain on Functional Expression Several T1 edits affected functional expression levels. In Figure 5C, gK was monitored for 4 days following cRNA injection using a 2-electrode voltage clamp. For all channel constructs, gK rose rapidly for the first 2 days and then leveled off. Expression levels for the channel constructs M35V, I43V, N45S, and K132R were not statistically different than for SqKv1.1G (not shown). The edits Y36C, R87G, K132E, and K132G, on the other hand, reduce functional expression. In the case of R87G, the reduction is dramatic: expression levels for this construct are only 3% that of SqKv1.1G (average for 4 days). For the other constructs, the reduction of maximal gK was intermediate. Expression levels for Y36C, K132G, and K132E channels are 64%, 66%, and 32% that of SqKv1.1G, respectively. Because expression levels can be strongly influenced by the cRNA sample’s integrity, cRNAs were resynthesized from freshly isolated plasmid DNA and the experiment was repeated using oocytes from a different frog (data not shown). In this experiment, overall expression levels were greater for each channel; however, the relative expression levels were similar to those described in the first experiment (61%, 58%, 33%, and 2% that of SqKv1.1G for Y36C, K132G, K132E, and R87G, respectively). In Vitro Binding Assay of T1 Domain Fusion Proteins Because of their position in the channel, we hypothesized that T1 edits, particularly those that reduce expression levels, modify the ␣ subunit’s ability to form tetramers. This possibility was explored by creating an in vitro binding assay using fusion proteins, synthesized in bacteria, which contain different versions of the T1 domain (Figure 6). The basic fusion protein construct contained the entire Sqkv1.1 T1 domain flanked by a single heart muscle kinase (K) motif on each end. A thrombin cleavage motif, preceded by a polyhistidine tag, were added on the N-terminal side. This T1 fusion

Figure 5. Electrophysiological Characterization of T1 Domain Edits (A and B) gK versus voltage relationship and ␶off versus voltage for T1 edit mutants. Experiments were performed in an analogous manner to Figure 4. (C) Characterization of expression levels for T1 edits. Maximum gK per oocyte (measured after a 25 ms test pulse to ⫹40 mV from a holding potential of ⫺80 mV) is plotted for 4 days following injection of cRNA (5 ng/oocyte). See Experimental Procedures for measurement of gK. All experiments performed at 20⬚C. Error bars (SEM) are included when they are bigger than the symbols. All constructs are significantly different from each other (Student’s t test; p ⱕ 0.05) except in the following cases: between K132G and Y36C on any day, between K132G and SqKv1.1G on day 1, between either K132G or Y36C and SqKv1.1G or K132E on day 4.

Neuron 750

Figure 6. T1 Binding Assay (A) Schematic of binding assay. T1 ⫽ amino acids 16–156 of SqKv1.1G. K ⫽ heart muscle kinase site. T ⫽ thrombin cleavage site. H ⫽ polyhistidine tag. NTA ⫽ nitrilotriacetic acid. Gray shading represents putative intermolecular bonding. (B) Assay products run on a 12% SDSPAGE gel and then exposed to film for 5 min. T1-G ⫽ unedited T1 target construct. T1-E ⫽ fully edited version. C-term is a fusion protein made to the C terminus (amino acids 414–488) of SqKv1.1G. N-term is made to the N terminus of SqKv1.1G (amino acids 87–215), and contains only half of the T1, an insufficient portion for binding. NC is made to the squid Na channel GFLN1 (amino acids 483–576; Rosenthal and Gilly, 1993). Unless otherwise indicated, 7.5 ␮g of target was added to the assay. Unlabeled T1-G probe was added as blocker. (C) A coomasie stain of the gel in (B) to verify that target constructs were loaded in proper amounts. An experiment similar to that in (B), but the assay products were counted in a liquid scintillation counter. Raw CPMs are blotted according to the assay’s target construct. The probe construct in all experiments was T1-G. All assays performed at 4⬚C.

construct constituted the binding “target” (Figure 6A). A “probe” was synthesized by cleaving target construct with thrombin and radiolabeling it with 32P at the kinase sites. The binding assay consisted of three steps (Figures 6Ai–6Aiii). First, the target was bound to Ni-NTA magnetic beads through its polyhistidine tag (Figure 6Ai). Then the radiolabeled probe construct, which lacked the polyhistidine tag, was added and allowed to bind to the target (Figure 6Aii) through T1 association. Finally the beads were washed and probe-target complexes were eluted with imidazole (Figure 6Aiii). Positive binding was assessed by analyzing products using either autoradiography following SDS-PAGE or liquid scintillation counting. Figure 6B shows an autoradiograph from several bind-

ing assay products run on a polyacrylamide gel. In this case, the basic T1 probe construct (unedited; T1-G) was allowed to bind to a variety of his-tagged target proteins. The basic T1 target (unedited; T1-G) clearly yields a robust band (lane 1). As negative controls, three separate fusion proteins yield only background binding (lanes 2–4). Binding no target to the magnetic beads yielded similar results (No target, lane 5). Signal intensity was dependent on the amount of T1-G target bound to the magnetic beads as a ten-fold reduction in target (lane 6) resulted in a weaker signal. The signal could also be reduced by the addition of unlabeled probe as blocker (lane 7). The addition of blocker, however, did not completely abolish the signal, a result expected from the extremely large target to probe ratio. Interestingly, the

RNA Editing Regulates K⫹ Channel Tetramerization 751

Table 1. Binding Assay Matrix for All T1 Fusion Protein Constructs Probe

Target T1-G T1-C M35V I43V N45S R87G K132E K132G K132R

T1-G

T1-C

M35V

2.5 ⫾ 0.20 ND 1.99 ⫾ 0.11 2.82 ⫾ 0.50 1.88 ⫾ 0.29 0.11 ⫾ 0.05 3.19 ⫾ 0.22 3.24 ⫾ 0.28 1.52 ⫾ 0.10

ND ND ND ND ND ND ND ND ND

2.26 ND 2.16 2.63 2.07 0.14 3.16 2.69 1.73

I43V

⫾ 0.21 ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾

0.36 0.33 0.22 0.03 0.22 0.26 0.21

2.23 ND 1.99 2.29 3.02 0.17 2.95 3.06 1.71

N45S ⫾ 0.43 ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾

0.28 0.32 0.40 0.03 0.45 0.44 0.30

1.17 ND 1.22 1.39 1.66 0.12 1.31 1.44 0.91

⫾ 0.05 ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾

0.10 0.16 0.24 0.03 0.04 0.02 0.08

R87G

K132E

ND ND ND ND ND ND ND ND ND

2.00 ND 2.24 2.14 2.39 0.41 2.09 2.39 1.78

K132G

⫾ 0.18 ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾

0.16 0.11 0.21 0.11 0.10 0.19 0.23

1.96 ND 1.82 2.02 2.46 0.58 1.81 1.96 1.69

K132R

⫾ 0.12 ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾

0.12 0.06 0.15 0.04 0.07 0.14 0.07

1.66 ND 1.88 1.87 2.29 0.32 1.78 1.98 1.65

⫾ 0.12 ⫾ ⫾ ⫾ ⫾ ⫾ ⫾ ⫾

0.11 0.26 0.25 0.02 0.19 0.16 0.14

Reported values are in ng of probe bound to 7.5 ␮g of target. As elsewhere, T1-G is the basic, unedited T1 fusion protein. Names of other constructs represent the indicated mutation in the T1-G background. T1-C has all T1 edits introduced. Errors ⫽ SD. Italic entries are highly different from normal (T1-G: T1-G) binding.

T1-E target, which was edited at all T1 editing sites, yielded no signal above background (lane 8). Figure 6C shows a coomasie blue stain of the same gel which produced the autoradiogram shown in Figure 6B. This verifies that approximately equal amounts of the target protein were bound to the magnetic beads in each binding assay. Figure 6D shows the results from a binding assay using identical target and probe constructs. In this case, the products were counted in a liquid scintillation counter instead of being run on a gel. Targets that lacked an entire T1 domain, as well as the no target control, all yielded a relatively small signal (ⵑ2,500 CPM). This level was considered nonspecific background. By contrast, the T1-G assay produced a robust signal of almost 40,000 CPM. As with the gel analysis, this level was reduced by either using less target or by the addition of cold probe as blocker. Again, the fully edited T1’s signal (T1-E) was indistinguishable from background. To test for changes in T1 association due to editing, our binding assay was extended to fusion proteins that contained single editing sites. These mutations (except for Y36C; see Experimental Procedures) were introduced into both probes and targets and all probe-target combinations were then tested. Table 1 presents the results from these experiments. Different trials were normalized by calculating the ng of probe bound per assay. In general, ⵑ1.5–3 ng of probe was bound. However, there were notable exceptions. Assays that contained the fully edited T1-E construct, whether probe or target, showed no binding. The R87G edit alone yielded a strong, negative effect. Using the R87G fusion construct as probe, no binding was detected. As a target, however, binding was evident, but at a greatly reduced level. Edits N45S and K132R also reduced binding, although to a much less dramatic degree. Position 132 has an interesting effect on binding when combined with the R87G edit. In general, most probe constructs bind poorly to the R87G target (0.11–0.17 ng). However, this amount can be significantly augmented by editing at position 132. In fact, the K132G edit increases the R87G signal ⵑ4-fold. These results demonstrate that several editing sites can affect T1 binding in our assay. However, of the sites that affect expression level, only R87G appears to alter T1 binding.

Sucrose Gradients to Measure Relative Oligomerization between T1-G and T1-G R87G Mechanistic interpretations based on our binding assay results are complicated by the fact that both target and probe may bind to themselves. Consequently, before the probe and target are mixed, they may already be present in an equilibrium between monomer, dimer, trimer, and tetramer. Thus, the differences in binding identified above could arise from different causes. For example, the reduction in the R87G signal could be due to a relatively low binding affinity. Conversely, the R87G construct could bind with a very high affinity to itself and to other T1 constructs. In this case, very little R87G probe would be available in the assay, resulting in a low signal. To distinguish between these possibilities, both the T1-G and the T1-G R87G fusion proteins were analyzed by ultracentrifugation on sucrose density gradients. Fusion protein samples were mixed with four molecular weight standards (known monomers), layered onto sucrose gradients, and subjected to ultracentrifugation. Fractions were collected from ultracentrifuge tube’s bottom, and a portion of each was run on SDS-PAGE gel. Figure 7A presents an example of an experiment using the T1-G fusion protein. As expected, the molecular weight standards migrate in the gradient according to their mass, with the largest proteins appearing in the earliest fractions. Bovine serum albumin (BSA), at 66 kDa, is detectable in the first fraction and peaks in ⵑfraction 6. On the other hand, cytochrome B (CB), at 12 kDa, does not appear until fraction 8 and peaks at about fraction 11. The T1-G monomer has a molecular weight of 20.2 kDa. If not self-associated into higher molecular weight forms, it would be expected to peak between CB and carbonic anyhdrase (CA; 29 kDa). This is clearly not the case. The T1-G signal appears in early fractions and is spread over a large portion of the gradient. The gel data was then quantified, and each band’s fractional intensity was plotted against it’s gradient fraction (Figure 7B). The signals for the standards form sharp peaks whose positions are determined by their molecular weights. The T1-G signal, however, is broad and centered between 29 kDa and 45 kDa, indicating that T1-G is partially associated into multimers. Furthermore, although it appears likely that the majority of T1-G proteins are present as either monomers or dimers, a small por-

Neuron 752

Figure 7. Sucrose Gradient Analysis of T1-G and T1-G R87G A linear 7.5%–27.5% sucrose gradient was loaded with 30 ␮g of T1-G (unlabeled probe construct) and 20 ␮g each of the following molecular weight standards: BSA (bovine serum albumin), EA (chicken egg albumin), CA (carbonic anhydrase), CB (cytochrome B). Fractions were collected from the bottom of the gradient and run on a 12% SDS-PAGE gel. (A) is a picture of that gel. (B) Quantification of results in (A). Each band’s fractional intensity, expressed as its intensity divided by the sum of the intensities for all bands for that protein in all fractions, was calculated. This fraction was then plotted versus fraction number. BSA ⫽ O, EA ⫽ 䊐, CA ⫽ ⌬, CB ⫽ 䉫, T1-G ⫽ 䊉. (C) An identical experiment to (A) and (B) except T1-G R87G was used instead of T1-G. T1-G R87G ⫽ 䊉. Thicker lines are used for T1-G and T1-G R87G.

tion of the signal is seen in substantially earlier fractions than the Egg Albumin standard (45 kDa), suggesting the presence of higher molecular weight forms as well. A similar density gradient analysis was extended to T1-G R87G fusion proteins (Figure 7C). In this case, the peak marking the R87G peptide had much the same shape as those of the standards. In addition, it was centered at the approximate position expected for the 20.2 kDa monomer. Based on this, it appears that the majority of the T1-G R87G construct is present as a monomer. We therefore conclude that T1-G R87G has a lower binding affinity than T1-G. Discussion The A→G transitions between genomic and cDNA sequences described in this report are consistent with RNA editing mediated by the hydrolytic deamination of A→I, and our general approach for identifying these sites has been followed before (Hanrahan et al., 2000; Patton et al., 1997; Petschek et al., 1996; Smith et al., 1996). Genomic Southern blot data, and the fact that no nucleotide sequence variability was ever encountered using genomic DNA as a PCR template, both argue strongly against the possibility that our data result from the pres-

ence of multiple SqKv1.1 genes. What is notable in the present case is the large number of editing sites identified and the wide variety of channel properties that they regulate. Using our assay, 14 sites were identified in SqKv1.1, and the editing frequency varied tremendously between positions. As with the squid Kv2 channel previously described (Patton et al., 1997), this density is very high compared to vertebrate examples. Assuming that editing at each site is independent, based on 13 editing sites that change codons, there are 8,192 possible SqKv1.1 ␣ subunit monomers and 4.5 ⫻ 1015 tetramers. Clearly, all permutations cannot be made, and editing at one site may well depend on editing at other sites, but the possibilities for regulation are extensive. In their seminal work, Hodgkin and Huxley clearly defined those biophysical properties of potassium conductance that are critical for action potential repolarization in nerve. Remarkably, RNA editing directly targets many of these properties. Not only are voltage dependence and voltage sensitivity effected, but the simplest parameter of all, absolute potassium conductance (gK), is modified in a novel way. Half of the editing sites in SqKv1.1 are located in sequence coding for the T1 domain. Recent investigations have revealed that the T1 domain is involved in more of the channel’s activities than previously suspected. For some time it has been known that the T1 helps restrict heteromultimer formation to compatible ␣ subunits (Deal et al., 1994; Li et al., 1992; Pfaffinger and DeRubeis, 1995; Schulteis et al., 1998; Shen and Pfaffinger, 1995; Tu et al., 1995; Xu et al., 1995). It also greatly improves the efficiency of tetramerization; however, in some cases, it is not absolutely necessary (Kobertz and Miller, 1999; Tu et al., 1995). Although sites of subunit interaction clearly exist in other channel domains (Sheng et al., 1997; Tu et al., 1996), the T1 is thought to drive tetramerization and greatly contribute to the overall stability of the mature channel structure (Strang et al., 2001). Because mutations of amino acids which line the intersubunit interaction surfaces cause shifts in voltage sensitivity, recent reports suggest that the T1 domain also influences gating (Cushman et al., 2000; Minor et al., 2000). RNA edits identified in this report alter both tetramerization and gating. According to the Aplysia T1 domain’s crystal structure, 17 amino acids participate in intersubunit polar interactions (Bixby et al., 1999; Kreusch et al., 1998). Of these, all but one are conserved in the squid genomic sequence. Two positions (N45S and R87G), however, can be edited. Of these, R87G is particularly noteworthy because it creates a large effect on channel expression and T1 binding. This position is located at a hinge between the T1’s second and third layers. It is interesting that position 87’s binding partner is the only T1 residue not conserved in squid (Q126 in Aplysia, H100 in squid). Why then does the R87G edit reduce T1 binding to such a degree? Several possibilities exist. First, D86, the adjacent residue, participates in five polar interactions with other residues. Perhaps editing position 87 disrupts bonding at position 86 or at other, more distant interaction points. It is also possible that the squid T1 structure differs from that of Aplysia, and position 87 has a large contribution to intersubunit bonding. The expression

RNA Editing Regulates K⫹ Channel Tetramerization 753

Table 2. Oligonucleotide Primers Used for This Study Name

Sequence

Position

Orientation

JR1 JR4 JR16 JR20 JR21 JR22 JR36 JR41 JR50 JR58 JR59 SKC8 GenLink1 GenLink2 Nest1 Nest2

cggatgaagagaacaaccag aaccactcccagcatcc aggctttaccgaaaacacaggac ttgctgcttccaaggtctct tgagatcagatctgtatcaaggccatagtctc cttccaaggtctcttgactt gttcaacgtaattgcca cccttgtacatatccacc gaacactgagcaggcaatgac agagaaccactcccagcatccaactgg gtccgcttacgttgatgataacacggtc caacgtgatgtttagaaaagttgtttca gtaatacgactcactatagggcacgcgtggtcgacggcccgggctggt P-gatcaccagccc-L gtaatacgactcactatagggc actatagggcacgcgtggt

56–75 77–94 (⫺53)–(⫺30) 22–41 1468–1487 28–47 (⫺152)–(⫺146) (⫺96)–(⫺79) 61–81 of intron 71–97 118–145 (⫺7)–14 NA NA NA NA

sense antisense sense sense antisense sense antisense sense antisense antisense antisense sense NA NA NA NA

Numbering refers to the SqKv1.1A cDNA. P ⫽ phosphate. L ⫽ camino linker.

level reduction due to R87G may also result from subtle subunit misfolding during biogenesis. Indeed, misfolding and reduced T1 binding may be related. Several examples exist where surface expression is greatly reduced by T1 mutations which influence proper protein folding (Manganas and Trimmer, 2000; Schulteis et al., 1998; Strang et al., 2001). The R87G mutation in the SqKv1.1A background significantly reduces total channel protein as determined by Western blots and also reduces expression level in oocytes (Liu et al., 2001). Other edits within the T1 domain lead to changes in functional properties. Both N45S and K132E shifted the g-V relationship to more depolarized potentials. N45 is the most highly conserved position in the entire T1 (Bixby et al., 1999) and is, in fact, the only residue absolutely conserved across all four Kv subfamilies. This residue also contributes to polar interactions and is located near the C terminus of the 1st ␤ sheet of layer 1. A mutation of this position in Kv1.2 (N38A) caused a similar shift in the g-V relationship (Minor et al., 2000). Position K132 resides on the outside face of the second ␣ helix of layer 3, a region that has not been the focus of previous analysis. Its effects on voltage sensitivity and expression level cannot be readily explained. Perhaps this face of the helix contacts the channel’s core, and mutations of position 132 affect the coupling between the T1 and the core gating apparatus. The functional consequences of editing are further complicated by the fact that K channels are tetramers. How are the effects of edits manifested in heteromultimers? This is particularly interesting in the case of R87G because it both drastically reduces expression level and is present at a high frequency. Can a single unedited subunit at this position (R87) rescue expression of the tetramer, or does R87G have a dominant negative effect? Conceivably, small changes in the editing frequency at this position could translate into large changes in expression level. The possibilities for regulation are even more extensive considering, in the T1 binding assay, edits at position K132 could increase R87G binding. However, constructs containing these edits on the same subunit (e.g., SqKv1.1G R87G K132E and SqKv1.1G R87G K132G, data not shown) did not yield

higher expression levels in oocytes than those containing the R87G edit alone. Experiments are presently being designed to examine the expression levels of heteromultimers containing these edits on different subunits. Other examples of a single subunit determining heteromultimer expression have been reported (Manganas and Trimmer, 2000; Panyi and Deutsch, 1996; Schulteis et al., 1998). More editing sites have now been identified in squid K⫹ channel mRNAs than in all vertebrate substrates combined. Consequently, this system offers an excellent opportunity to question the underlying purpose of this process. First, we must note that there is only one silent editing site in SqKv1.1. In SqKv 2, 5 of 18 edits are silent; however, they are all edited at low efficiencies. This suggests that in squid, the process is specifically targeting protein function. Comparisons between the specific effects of editing on SqKv1.1 and SqKv 2 function are rendered difficult by the fact that only a small portion of SqKv 2’s core (S4-S6) has thus far been examined. As with SqKv1.1, there are many conversions to valine within SqKv 2’s membrane spans. In addition, one edit, SqKv 2 I597V, also speeds closing kinetics. It will be interesting to extend the analysis of SqKv 2 to the T1 domain. A notable difference between the two channels is the relatively high density of editing sites in SqKv 2. However, in SqKv 2, where editing sites were identified by sequencing individual clones, over half the sites were edited at low frequencies. Thus, more low frequency sites could be uncovered in SqKv1.1 by sequencing individual isolates. Why edit K⫹ channels? The most straightforward explanation is that editing is directed specifically at K⫹ channel function. Here gK and voltage sensitivity can be regulated in incremental steps. Different combinations of editing sites can determine the amount of gK that is recruited during an action potential, and the speed at which this conductance turns off. Consequently the duration of the action potential’s falling phase and afterhyperpolarization can be regulated. This argument, however, assumes that functional changes measured in Xenopus oocytes accurately mirror those in squid neurons. This assumption will have to be tested. Another

Neuron 754

possibility stems from the observation that all editing sites in SqKv1.1G that have functional consequences decrease available gK (with the modest exception of I387V) . This is accomplished either by changing functional expression, shifting the g-V relationship to more depolarized potentials, or speeding closing kinetics. It is therefore possible that RNA editing is slowly silencing SqKv1.1G’s function. A third intriguing possibility is that the functional effects of RNA editing are not tailored specifically for SqKv1.1. Instead, editing causes common effects among all proteins. Thus far in squid, 26 amino acids can be changed by editing. Of these, 22 result in the substitution of a smaller residue. In fact, there is only one case of a modest increase in amino acid size (SqKv1.1G K132R) and this mutation results from two edits within the same codon. In 12 cases, edits result in valine, and in four cases, they result in glycine. Within membrane spans, 8 out of a total of 11 edits result in V. This pattern has a striking resemblance to mutations arising from cold temperature adaptation in other proteins. For example, different lactate dehydrogenases isolated from Antarctic fish have small amino acids substituted at hinges and contact points (Fields and Somero, 1998). The same is true for closely related barracudas that inhabit different thermal environments (Holland et al., 1997). In these cases, mutations are thought to increase flexibility and entropy to compensate for the cold. Although these studies have not been extended to membrane proteins, the propensity for V substitutions at the putative membrane interface could have similar consequences. R87G is clearly situated at a “hinge” in the T1 domain. The L. opalescense used for this study live approximately between 7⬚C –14⬚C. This is substantially colder than for other organisms from which K⫹ channels have been isolated. Do organisms that live at cold temperatures edit more? The identification of more editing sites, in different mRNA substrates, and in different organisms, will help shed light on these possibilities.

diluted 500-fold and reamplified using a nested primer pair. Products (pSF) were cloned and sequenced using standard techniques (Sambrook et al., 1989). Sequence from these reactions confirmed the first 35 nt of the SqKv1.1A coding sequence. After this point, however, the sequence differed dramatically from SqKv1.1A. Thus far, 163 nt of 5⬘ intron sequence have been isolated. The sequence for pSF was the consensus from three individual products. Adaptor-Mediated Genomic Walking The 3⬘ side of the intron was isolated by a linker mediated genomic walk using BstYI fragments of genomic DNA, and nested PCR primer pairs JR59&Nest1 and JR58&Nest2. PCR primer sites that flanked DNA target sequences were created by ligating adaptor oligonucleotides to 2.5 ␮g of genomic DNA digested with BstYI. 0.5 ␮g of this product was ligated to 50 pmol of adaptor with T4 DNA ligase in a total volume of 8 ␮l. One microliter of this reaction was used for amplification with primer “Nest1” and a gene-specific primer (JR59). This reaction was then subjected to further amplification with primers Nest2 and a nested gene-specific primer (JR58). The adaptor primer was made by annealing 1 nmol of the oligonucleotide GenLink1 with 1 nmol of GenLink2. Primer GenLink1 contains target sites for Nest1 and Nest2. Oligonucleotide Genlink2 contained a 5⬘ phosphate for ligation and a 3⬘ amino linker to block extension. In this way, only products extended by the gene-specific primers could serve as templates for further amplification. Reaction products were cloned, and sequence from these products was highly similar to SqKv1.1A after nt 36. The 731 nt of sequence obtained before this point, however, showed no similarity. Southern Blots Genomic DNA was isolated from a Loligo opalescense collected from Santa Monica Bay. A portion of the squid’s sperm sack was pulverized in a mortar and pestle after being frozen with liquid nitrogen and high molecular weight DNA was isolated using genomictip 100/G columns (Qiagen) according to the supplied instructions. Due to the abundance of DNA in sperm, only 1/8th of the suggested starting material was used. Southern blots were performed according to the instructions supplied with Hybond NX transfer membranes (Amersham). For each lane, 10 ␮g of genomic DNA was digested with the appropriate restriction enzyme and run on a 0.7% agarose gel overnight at low voltage. Hybridization and washing were performed under high stringency conditions. 32P-labeled probe was generated by random priming (Sambrook et al., 1989) using SqKv1.1G restriction fragments (see Figure 1 legend) and added at a concentration of 1 ⫻ 106 CPM/ml. Blots were exposed to X-ray film for 2–5 days.

Experimental Procedures Isolation and Characterization of Genomic DNA for SqKv1.1A PCR to Test for the Presence of an Intron The prediction that an intron was present in SqKv1.1A genomic DNA between nt 35-36 was supported by several PCR trials using squid genomic DNA as template and primers that spanned this position. Amplifications using sense primers JR41, JR16, SKC 8, or JR20 and antisense primers JR4, JR58, JR59, or JR21, in any combination, were unsuccessful (see Table 2 for the position and sequence of all oligonucleotides). However, amplifications using primer pairs on either side of the putative intron were successful and yielded products of the size predicted from the cDNA sequence. An amplification using sense primer JR1, which is on the 3⬘ side of this position, and antisense primer JR21, which is in the 3⬘ utr yielded a product (p121) of the expected size. This clone contained nearly the entire SqKv1.1A coding sequence and had no introns. Inverse PCR The 5⬘ side of the intron was isolated by inverse PCR (Triglia et al., 1988) from circularized Sau3A fragments of squid genomic DNA using the primer pair JR36 and JR41 (Figure 1A). For a single reaction, 200 ng of genomic DNA was digested to completion with Sau3A (New England Biolabs). To make circular DNA, samples, at a concentration of 5 ng/␮l, were ligated with T4 DNA ligase (New England Biolabs) at 15⬚C overnight. Samples were then amplified using Pfu DNA polymerase (Stratagene). Initial amplification products were

Direct Sequencing Assay for RNA Editing Our RNA editing assay consisted of four steps: (1) isolate genomic DNA and Giant Fiber Lobe (GFL) total RNA from an individual squid, (2) make cDNA from the total RNA, (3) amplify SqKv1.1 from both sources, (4) directly sequence both products using 33P-labeled ddNTPs. Genomic DNA was isolated from a small piece of mantle tissue using Genomic-tip 20/G columns (Qiagen) according to the suggested protocol. Total RNA was isolated from single giant fiber lobes (GFL). The GFL was homogenized in 800 ul of RNAzol B (AMS Biotechnology, Oxon, UK) in a frosted glass homogenizer. The RNA was then isolated according to the supplied instructions. As a carrier, 5 ␮g of glycogen was added prior to the isopropanol precipitation. RNA was treated with DNaseI (BRL). The RNA pellet was resuspended in 10 ul 1 mM EDTA, and split evenly into two tubes. The contents of one tube were then used as template for first strand cDNA synthesis using a Superscript reverse transcriptase kit (Bethesda Research Laboratories) according to the supplied instructions (Oligo dT was used as a primer; RNaseH tratment step included). The second tube of RNA was used as a no reverse transcript control to test for genomic DNA contamination. Either 1/5th the cDNA synthesis reaction, or 200 ng of genomic DNA, was used as a template to amplify SqKv1.1 using Pfu DNA polymerase (Stratagene). For genomic DNA, primers JR41 and JR50 were used to amplify exon 1, and JR20 and JR21 were used to amplify exon 2. For cDNA, primers JR16 and JR4 were used to amplify sequence encoded by exon 1, and JR21 and JR22 were

RNA Editing Regulates K⫹ Channel Tetramerization 755

used to amplify sequence encoded by exon 2. PCR amplification used 35 cycles. Each cycle consisted of 45 s of denaturation at 95⬚C, 45 s of annealing at 63⬚C, and 3 min of extension at 72⬚C. The PCR reaction contained 500 nM of each primer, template as indicated above, 250 ␮M dNTPs, and 2.5 units Pfu in a total volume of 25 ␮l. Products were gel purified with spin columns (Qiagen) and eluted into 30 ␮l water. Three microliters was then used as template for direct sequencing using a Thermo Sequenase 33P-Radiolabeled Terminator Cycle Sequencing Kit (Amersham) according to instructions. Sequencing primers were made at 100 bp intervals along the SqKv1.1 sequence. Reaction products were loaded onto a 6% sequencing gel and run at 60 W using a glycerol tolerant running buffer (90 mM Tris, 29 mM taurine, 1 mM ethylenediaminetetracetic acid). After running, gels were dried and read on a phosphorimager. Band intensities were quantified with ImageQuant software by drawing a 10 pixel wide line through the center of each band and integrating the resulting peak’s intensity. Because this assay was based on direct sequencing, random PCR errors were not a problem. In no case were “false” editing sites encountered while using plasmid controls or genomic DNA as PCR template. Functional Expression in Xenopus Oocytes Constructs for functional channel expression were derived from the cDNA clone SqKv1.1A (Chi 7-pBSTA; Rosenthal et al., 1996). This construct contains the SqKv1A cDNA coding region cloned between the Xenopus ␤-globin 5⬘ and 3⬘ untranslated regions. To make a non-edited version of this construct, PCR primers JR20 and JR22 were used to amplify SqKv1.1 from genomic DNA. This product was cut with restriction enzymes NdeI and SpeI and cloned into the equivalent sites of Chi7-pBSTA to make SqKv1.1G. Mutations were introduced into SqKv1.1G by using the Quik Change Site Directed Mutagenesis kit (Stratagene) or by subcloning restriction fragments from SqKv1A. All mutants were verified by DNA sequencing. After linearizing these plasmids with NotI, cRNA for oocyte expression was transcribed using the Message Machine Kit (Ambion). Oocytes were isolated and injected as previously described (Rosenthal et al., 1996). Each oocyte was injected with 5 ng cRNA. Channel currents were measured from oocytes 2 days after cRNA injection using the cut-open vaseline gap method (COVG) (Stefani and Bezanilla, 1998). Analog signals were filtered at 1/10th the sampling rate, digitized with a PC44 board (Innovative Integration, Simi Valley, CA) and collected using software written in our lab. Linear leak currents, and membrane capacitive currents, were subtracted using a standard online P/4 procedure. To gain electrical access to the oocyte’s interior, Nystatin (20 ␮g/ml) was used instead of saponin. This practice greatly reduced problems with current rundown over the course of the experiment. To minimize errors caused by series resistance, oocytes expressing currents greater than 10 ␮A at strong depolarizations were excluded from analysis. The largest series resistance errors were estimated to be less than 6 mV. External solution consisted of (in mM) 20 K⫹-methansesulfonic acid (MES), 100 N-methyl glucamine (NMG)-MES, 2 CaCl2, 10 HEPES, pH 7.4. Internal solution contained 120 K⫹-MES, 2 EGTA, 10 HEPES, pH 7.4. Expression level studies were conducted using a Geneclamp 500B 2 electrode voltage clamp (Axon Instruments). Signal processing was performed as described above. Data were collected using Vclamp software (Rosenthal et al., 1996). Electrodes were ⵑ0.3–0.6 M⍀. The external solution was the same as that used in the previous section. For all experiments using oocytes, the temperature was maintained at 20⬚C and oocytes were held at ⫺80 mV. Data Analysis The gK versus voltage relationship was computed by measuring peak tail currents following test pulses to various potentials from a holding potential of ⫺80 mV. These measurements were then normalized to the maximal value. Deactivation kinetics were measured by repolarizing the oocyte to various negative values following an activating pulse to ⫹50 mV for 10 ms. Tail currents were fit to a single exponential of the form y ⫽ Aexp(⫺␶off/T) ⫹ B where A is the maximum amplitude, T is time, B is the baseline, and ␶off is the time constant. To measure maximum gK for expression level studies, oocytes were pulsed to ⱖ30 mV for 25 ms, and then returned to the

holding potential of ⫺80 mV. The instantaneous current magnitude (outward current ⫹ |tail current|) at 25 ms was then divided by the instantaneous change in voltage. Efforts were made to select the smallest current record that was still in the nearly saturated range of the conductance versus voltage relationship (Vm ⱖ 30 mV). In spite of this, series resistance errors were significant for constructs that expressed at high levels (e.g., SqKv1.1G). Expression levels for these constructs were undoubtedly much higher than those reported. These errors, however, in no way affect our conclusions. Generation of Fusion Proteins All fusion proteins were made in the pET 15B expression vector (Novagen). For T1-G, oligonucleotide primers JR72 and JR73 were used to amplify nt 49-466 of SqKv1.1G (amino acids K17-R155). These primers contain sequence encoding heart muscle kinase sites (RRASV) and BamHI restriction sites flanking the squid sequence. Products were cloned into the BamHI site of pET 15B. Other T1 fusion proteins were made in an analogous manner except that expression constructs, which contained editing site point mutations, were substituted as template for PCR. All reactions used Pfu DNA polymerase. The integrity of all constructs was verified by sequencing. The bacterial strain Bl21 DE3 (Novagen) was used to generate fusion proteins. His-tagged proteins were isolated from 100 ml cultures induced with 1 mM IPTG, using Ni-Nitrilotriacetic acid (NTA) Agarose (Qiagen) following the protocol supplied for denaturing conditions. For refolding, the sample was dialyzed against many 1 liter changes of lysis buffer pH 8 (8 M Urea, 0.1M NaH2PO4, 10 mM Tris, pH8) for 2 hr at 4⬚C. Each change contained 1 M less Urea. Changes were repeated until no Urea remained. Protein concentration was determined by a BCA assay (Pierce). The His-tag was removed from a portion of each protein preparation using the Thrombin Cleavage Capture Kit (Novagen) according to supplied instructions. T1 Binding Assay Although it contains numerous modifications, our binding assay is based on methods outlined in Xu and Li (1998). To synthesize probe, the basic labeling reaction contained: 3 ␮g thrombin cleaved fusion protein, 12 ␮l 10⫻ HKE buffer (in mM: 200 mM Tris, 10 mM DTT, 1000 NaCl, 120 mM MgCl2, pH 7.8), 4 ␮l heart muscle kinase (Sigma P2645, resuspended in 40 mM DTT to a concentration of 10 U/␮l), H20 to a final volume of 120 ␮l. The reaction was incubated at 37⬚C for 90 min, diluted to 1 ml with dialysis buffer (in mM: 60 KCl, 10 HEPES, 1 EDTA, pH 7.5), and dialyzed against 500 ml of the same buffer (4 buffer changes over a 12 hr period). Specific activity was determined by a standard TCA precipitation and was routinely 0.5–1 ⫻ 106 CPM/pmol. The basic assay consisted of binding 7.5 ␮g of His-tagged fusion protein to 20 ␮l of Ni-NTA magnetic agarose beads (Qiagen) in 500 ␮l of binding buffer (in mM: 50 NaH2PO4, 300 NaCl, 20 Imidazole, pH 7.5) for 1 hr at room temperature. Using a magnetic stand, the beads were then washed once with 1 ml of binding buffer ⫹ 0.005% Tween 20, and resuspended in 400 ␮l interaction buffer (in mM: 50 NaH2PO4, 100 KCl, 20 Imidazole, 0.005% Tween 20, 0.2% BSA, pH 7.5). Twenty microliters of 32P-labeled probe was added and the reaction was incubated for 2 hr at 4⬚C. Beads were then washed with 1 ml ice cold interaction buffer and probe-target complexes were eluted with 50 ␮l of elution buffer (in mM: 50 NaH2PO4, 100 KCl, 250 Imidazole, 0.005% Tween 20, 0.2% BSA, pH 7.5). Eluate was then either analyzed by autoradiography following SDS-PAGE, or counted in a liquid scintillation counter. The Y36C fusion protein was not analyzed in the assay because it formed disulfide bonds under the above conditions. Sucrose Gradients All sucrose solutions were made in gradient buffer (in mM: 50 NaH2PO4, 100 NaCl, 10 HEPES, pH 7). Gradients, consisting of 27.5% to 7.5% sucrose in 2.5% steps, were made by layering 200 ␮l of each sucrose solution in a Beckman pollyallomer (cat #: 347357) ultracentrifuge tube. Gradients were then frozen and allowed to thaw. Once thawed, 20 ␮g of each standard, and 30 ␮g of the relevant T1 fusion protein, were layered on top. Ultracentrifugation

Neuron 756

was carried out at 55,000 RPM for 15 hr at 2⬚C using a TLS-55 swinging bucket rotor in a Beckman Optima TL ultracentrifuge. At the end of the run, 100 ␮l fractions were collected from the bottom of the tube and 20 ␮l of each was run on a 12% SDS-PAGE gel. Gels were stained with Gelcode Blue Stain Reagent (Pierce), dried, and scanned. Bands were quantified using Scion Image software (Scion Corporation, Frederick, MD) by drawing rectangles which fully enclosed each band. Average pixel intensity within these rectangles was calculated after background intensity was subtracted. Molecular weight monomer standards were: bovine serum albumin (66 kDa), egg albumin (45 kDa), carbonic anhydrase (29 kDa), and cytochrome B (12 kDa). Acknowledgments

(1999). RNA editing of a Drosophila sodium channel gene. Ann. NY Acad. Sci. 868, 51–66. Hanrahan, C.J., Palladino, M.J., Ganetzky, B., and Reenan, R.A. (2000). RNA editing of the Drosophila para Na(⫹) channel transcript. Evolutionary conservation and developmental regulation. Genetics 155, 1149–1160. Hodgkin, A.L., and Huxley, A.F. (1952). A quantitative description of membrane current and its application to conduction and excitation in nerve. J. Physiol. 117, 500–541. Holland, L.Z., McFall-Ngai, M., and Somero, G.N. (1997). Evolution of lactate dehydrogenase-A homologs of barracuda fishes (genus Sphyraena) from different thermal environments: differences in kinetic properties and thermal stability are due to amino acid substitutions outside the active site. Biochemistry 36, 3207–3215.

We wish to thank Dr. William Gilly for generously providing animals and laboratory facilities. We also thank Dr. Miguel Holmgren and Dr. Dorine Starace for critically reviewing the manuscript. This work was supported by NRSA GM20314 to J.R. and NIH grant GM30376 to F.B.

Hong, K.H., and Miller, C. (2000). The lipid-protein interface of a Shaker K(⫹) channel. J. Gen. Physiol. 115, 51–58.

Received: December 29, 2001 Revised: March 18, 2002

Kobertz, W.R., and Miller, C. (1999). K⫹ channels lacking the ‘tetramerization’ domain: implications for pore structure. Nat. Struct. Biol. 6, 1122–1125.

References

Kohler, M., Burnashev, N., Sakmann, B., and Seeburg, P.H. (1993). Determinants of Ca2⫹ permeability in both TM1 and TM2 of high affinity kainate receptor channels: diversity by RNA editing. Neuron 10, 491–500.

Armstrong, C.M., and Bezanilla, F. (1973). Currents related to movement of the gating particles of the sodium channels. Nature 242, 459–461.

Kim, U., Wang, Y., Sanford, T., Zeng, Y., and Nishikura, K. (1994). Molecular cloning of cDNA for double-stranded RNA adenosine deaminase, a candidate enzyme for nuclear RNA editing. Proc. Natl. Acad. Sci. USA 91, 11457–11461.

Bass, B.L., and Weintraub, H. (1987). A developmentally regulated activity that unwinds RNA duplexes. Cell 48, 607–613.

Kreusch, A., Pfaffinger, P.J., Stevens, C.F., and Choe, S. (1998). Crystal structure of the tetramerization domain of the Shaker potassium channel. Nature 392, 945–948.

Bass, B.L., and Weintraub, H. (1988). An unwinding activity that covalently modifies its double-stranded RNA substrate. Cell 55, 1089–1098.

Li, M., Jan, Y.N., and Jan, L.Y. (1992). Specification of subunit assembly by the hydrophilic amino-terminal domain of the Shaker potassium channel. Science 257, 1225–1230.

Bass, B.L., Weintraub, H., Cattaneo, R., and Billeter, M.A. (1989). Biased hypermutation of viral RNA genomes could be due to unwinding/modification of double-stranded RNA. Cell 56, 331.

Li, M., Isacoff, E., Jan, Y.N., and Jan, L.Y. (1993). Assembly of potassium channels. Ann. NY Acad. Sci. 707, 51–59.

Bezanilla, F. (2000). The voltage sensor in voltage-dependent ion channels. Physiol. Rev. 80, 555–592. Bezanilla, F., Rojas, E., and Taylor, R.E. (1970a). Sodium and potassium conductance changes during a membrane action potential. J. Physiol. 211, 729–751. Bezanilla, F., Rojas, E., and Taylor, R.E. (1970b). Time course of the sodium influx in squid giant axon during a single voltage clamp pulse. J. Physiol. 207, 151–164. Bixby, K.A., Nanao, M.H., Shen, N.V., Kreusch, A., Bellamy, H., Pfaffinger, P.J., and Choe, S. (1999). Zn2⫹-binding and molecular determinants of tetramerization in voltage-gated K⫹ channels. Nat. Struct. Biol. 6, 38–43. Burns, C.M., Chu, H., Rueter, S.M., Hutchinson, L.K., Canton, H., Sanders-Bush, E., and Emeson, R.B. (1997). Regulation of serotonin2C receptor G-protein coupling by RNA editing. Nature 387, 303–308. Cushman, S.J., Nanao, M.H., Jahng, A.W., DeRubeis, D., Choe, S., and Pfaffinger, P.J. (2000). Voltage dependent activation of potassium channels is coupled to T1 domain structure. Nat. Struct. Biol. 7, 403–407. Deal, K.K., Lovinger, D.M., and Tamkun, M.M. (1994). The brain Kv1.1 potassium channel: in vitro and in vivo studies on subunit assembly and posttranslational processing. J. Neurosci. 14, 1666– 1676. Doyle, D.A., Morais Cabral, J., Pfuetzner, R.A., Kuo, A., Gulbis, J.M., Cohen, S.L., Chait, B.T., and MacKinnon, R. (1998). The structure of the potassium channel: molecular basis of K⫹ conduction and selectivity. Science 280, 69–77.

Li-Smerin, Y., Hackos, D.H., and Swartz, K.J. (2000). alpha-helical structural elements within the voltage-sensing domains of a K(⫹) channel. J. Gen. Physiol. 115, 33–50. Liu, T.I., Lebaric, Z.N., Rosenthal, J.J.C., and Gilly, W.F. (2001). Natural substitutions at highly conserved T1-domain residues perturb processing and functional expression of squid Kv1 channels. J. Neurophysiol. 85, 61–71. Llano, I., Webb, C.K., and Bezanilla, F. (1988). Potassium conductance of the squid giant axon. Single-channel studies. J. Gen. Physiol. 92, 179–196. Lomeli, H., Mosbacher, J., Melcher, T., Hoger, T., Geiger, J.R., Kuner, T., Monyer, H., Higuchi, M., Bach, A., and Seeburg, P.H. (1994). Control of kinetic properties of AMPA receptor channels by nuclear RNA editing. Science 266, 1709–1713. MacKinnon, R. (1991). Determination of the subunit stoichiometry of a voltage-activated potassium channel. Nature 350, 232–235. Manganas, L.N., and Trimmer, J.S. (2000). Subunit composition determines Kv1 potassium channel surface expression. J. Biol. Chem. 275, 29685–29693. Melcher, T., Maas, S., Herb, A., Sprengel, R., Higuchi, M., and Seeburg, P.H. (1996a). RED2, a brain-specific member of the RNAspecific adenosine deaminase family. J. Biol. Chem. 271, 31795– 31798. Melcher, T., Maas, S., Herb, A., Sprengel, R., Seeburg, P.H., and Higuchi, M. (1996b). A mammalian RNA editing enzyme. Nature 379, 460–464. Minor, D.L., Lin, Y.F., Mobley, B.C., Avelar, A., Jan, Y.N., Jan, L.Y., and Berger, J.M. (2000). The polar T1 interface is linked to conformational changes that open the voltage-gated potassium channel. Cell 102, 657–670.

Fields, P.A., and Somero, G.N. (1998). Hot spots in cold adaptation: localized increases in conformational flexibility in lactate dehydrogenase A4 orthologs of Antarctic notothenioid fishes. Proc. Natl. Acad. Sci. USA 95, 11476–11481.

Morais-Cabral, J.H., Zhou, Y., and MacKinnon, R. (2001). Energetic optimization of ion conduction rate by the K⫹ selectivity filter. Nature 414, 37–42.

Hanrahan, C.J., Palladino, M.J., Bonneau, L.J., and Reenan, R.A.

O’Connell, M.A., Krause, S., Higuchi, M., Hsuan, J.J., Totty, N.F.,

RNA Editing Regulates K⫹ Channel Tetramerization 757

Jenny, A., and Keller, W. (1995). Cloning of cDNAs encoding mammalian double-stranded RNA-specific adenosine deaminase. Mol. Cell. Biol. 15, 1389–1397. O’Connell, M.A., Gerber, A., and Keller, W. (1997). Purification of human double-stranded RNA-specific editase 1 (hRED1) involved in editing of brain glutamate receptor B pre-mRNA. J. Biol. Chem. 272, 473–478. Padgett, R.A., Grabowski, P.J., Konarska, M.M., Seiler, S., and Sharp, P.A. (1986). Splicing of messenger RNA precursors. Annu. Rev. Biochem. 55, 1119–1150. Panyi, G., and Deutsch, C. (1996). Assembly and suppression of endogenous Kv1.3 channels in human T cells. J. Gen. Physiol. 107, 409–420. Patton, D.E., Silva, T., and Bezanilla, F. (1997). RNA editing generates a diverse array of transcripts encoding squid Kv2 K⫹ channels with altered functional properties. Neuron 19, 711–722. Paul, M.S., and Bass, B.L. (1998). Inosine exists in mRNA at tissuespecific levels and is most abundant in brain mRNA. EMBO J. 17, 1120–1127. Petschek, J.P., Mermer, M.J., Scheckelhoff, M.R., Simone, A.A., and Vaughn, J.C. (1996). RNA editing in Drosophila 4f-rnp gene nuclear transcripts by multiple A-to-G conversions. J. Mol. Biol. 259, 885–890.

maps to a genetic locus associated with behavioral and visual defects. J. Neurosci. 16, 7868–7879. Smith, L.A., Peixoto, A.A., and Hall, J.C. (1998). RNA editing in the Drosophila DMCA1A calcium-channel alpha 1 subunit transcript. J. Neurogenet. 12, 227–240. Sommer, B., Kohler, M., Sprengel, R., and Seeburg, P.H. (1991). RNA editing in brain controls a determinant of ion flow in glutamategated channels. Cell 67, 11–19. Stefani, E., and Bezanilla, F. (1998). Cut-open oocyte voltage-clamp technique. Methods Enzymol. 293, 300–318. Strang, C., Cushman, S.J., DeRubeis, D., Peterson, D., and Pfaffinger, P.J. (2001). A central role for the T1 domain in voltage-gated potassium channel formation and function. J. Biol. Chem. 276, 28493–28502. Triglia, T., Peterson, M.G., and Kemp, D.J. (1988). A procedure for in vitro amplification of DNA segments that lie outside the boundaries of known sequences. Nucleic Acids Res. 16, 8186. Tu, L., Santarelli, V., and Deutsch, C. (1995). Truncated K⫹ channel DNA sequences specifically suppress lymphocyte K⫹ channel gene expression. Biophys. J. 68, 147–156. Tu, L., Santarelli, V., Sheng, Z., Skach, W., Pain, D., and Deutsch, C. (1996). Voltage-gated K⫹ channels contain multiple intersubunit association sites. J. Biol. Chem. 271, 18904–18911.

Petschek, J.P., Scheckelhoff, M.R., Mermer, M.J., and Vaughn, J.C. (1997). RNA editing and alternative splicing generate mRNA transcript diversity from the Drosophila 4f-rnp locus. Gene 204, 267–276.

Vandenberg, C.A., and Bezanilla, F. (1991). Single-channel, macroscopic, and gating currents from sodium channels in the squid giant axon. Biophys. J. 60, 1499–1510.

Pfaffinger, P.J., and DeRubeis, D. (1995). Shaker K⫹ channel T1 domain self-tetramerizes to a stable structure. J. Biol. Chem. 270, 28595–28600.

Xu, J., and Li, M. (1998). Methods used to study subunit assembly of potassium channels. Methods Enzymol. 293, 3–17.

Polson, A.G., Crain, P.F., Pomerantz, S.C., McCloskey, J.A., and Bass, B.L. (1991). The mechanism of adenosine to inosine conversion by the double-stranded RNA unwinding/modifying activity: a high-performance liquid chromatography-mass spectrometry analysis. Biochemistry 30, 11507–11514. Polson, A.G., Bass, B.L., and Casey, J.L. (1996). RNA editing of hepatitis delta virus antigenome by dsRNA-adenosine deaminase. Nature 380, 454–456. Pongs, O. (1999). Voltage-gated potassium channels: from hyperexcitability to excitement. FEBS Lett. 452, 31–35. Rosenthal, J.J., and Gilly, W.F. (1993). Amino acid sequence of a putative sodium channel expressed in the giant axon of the squid Loligo opalescens. Proc. Natl. Acad. Sci. USA 90, 10026–10030. Rosenthal, J.J., Vickery, R.G., and Gilly, W.F. (1996). Molecular identification of SqKv1A. A candidate for the delayed rectifier K channel in squid giant axon. J. Gen. Physiol. 108, 207–219. Rosenthal, J.J., Liu, T.I., and Gilly, W.F. (1997). A family of delayed rectifier Kv1 cDNAs showing cell type-specific expression in the squid stellate ganglion/giant fiber lobe complex. J. Neurosci. 17, 5070–5079. Sambrook, J., Fritsch, E.F., and Maniatis, T. (1989). Molecular Cloning: A Laboratory Manual (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory Press). Schulteis, C.T., Nagaya, N., and Papazian, D.M. (1998). Subunit folding and assembly steps are interspersed during Shaker potassium channel biogenesis. J. Biol. Chem. 273, 26210–26217. Semenov, E.P., and Pak, W.L. (1999). Diversification of Drosophila chloride channel gene by multiple posttranscriptional mRNA modifications. J. Neurochem. 72, 66–72. Shen, N.V., and Pfaffinger, P.J. (1995). Molecular recognition and assembly sequences involved in the subfamily-specific assembly of voltage-gated K⫹ channel subunit proteins. Neuron 14, 625–633. Shen, N.V., Chen, X., Boyer, M.M., and Pfaffinger, P.J. (1993). Deletion analysis of K⫹ channel assembly. Neuron 11, 67–76. Sheng, Z., Skach, W., Santarelli, V., and Deutsch, C. (1997). Evidence for interaction between transmembrane segments in assembly of Kv1.3. Biochemistry 36, 15501–15513. Smith, L.A., Wang, X., Peixoto, A.A., Neumann, E.K., Hall, L.M., and Hall, J.C. (1996). A Drosophila calcium channel alpha1 subunit gene

Xu, J., Yu, W., Jan, Y.N., Jan, L.Y., and Li, M. (1995). Assembly of voltage-gated potassium channels. Conserved hydrophilic motifs determine subfamily-specific interactions between the alpha-subunits. J. Biol. Chem. 270, 24761–24768.